VRDeepSafety: A Scalable VR Simulation Platform with V2X Communication for Enhanced Accident Prediction in Autonomous Vehicles

World Electric Vehicle Journal, Journal Year: 2025, Volume and Issue: 16(2), P. 82 - 82

Published: Feb. 6, 2025

Safe real-world navigation for autonomous vehicles (AVs) requires robust perception and decision-making, especially in complex, multi-agent scenarios. Existing AV datasets are limited by their inability to capture diverse V2X communication scenarios, lack of synchronized multi-sensor data, insufficient coverage critical edge cases multi-vehicle interactions. This paper introduces VRDeepSafety, a novel scalable VR simulation platform that overcomes these limitations integrating Vehicle-to-Everything (V2X) communication, including realistic latency, packet loss, signal prioritization, enhance accident prediction mitigation. VRDeepSafety generates comprehensive featuring interactions, coordinated sensor visual, LiDAR, radar, streams. Evaluated with our deep-learning model, VRFormer, which uniquely fuses data using probabilistic Bayesian inference, as well hierarchical Kalman particle filter structure, achieved an 85% accuracy (APA) at 2 s horizon, 17% increase 3D object detection precision (mAP), 0.3 reduction response time, outperforming single-vehicle baseline. Furthermore, integration increased APA 15%. Extending the horizon 3–4 reduced 70%, highlighting trade-off between time accuracy. The high-fidelity integrated provide valuable rigorous tool developing safer more responsive AVs.

Language: Английский

Robustness Verification of a Reinforcement Learning-Based Agent for Automated Car Parking

Lecture notes in electrical engineering, Journal Year: 2025, Volume and Issue: unknown, P. 139 - 147

Published: Jan. 1, 2025

Language: Английский